Lyria 3.5 for Text to Music

ABOUT THE WORKFLOW

Generate a Music Track
Describe a song or soundtrack in a prompt and get a full audio track back. Set the genre, instruments, tempo, and structure in the prompt. Add an optional image to match the music to a scene. The model returns stereo audio and lyrics as separate outputs.

Partner node. This workflow calls an external API, so each run uses credits from your API wallet. No API key needed. Floyo handles the connection.

Model

• Lyria 3.5 by Google DeepMind. A music generation model that writes and performs full songs with expressive vocals, rich arrangements, and 44.1 kHz stereo output.

HOW IT WORKS

Step 1. Write your prompt
Describe what you want to hear. Include genre, instruments, tempo, key, and song structure. The more specific, the better the result.
Works great with: orchestral scores · pop songs · ambient soundtracks · lo-fi beats

Step 2. Add an image (optional)
Upload a photo and the model writes music matching the mood and scene. Leave empty for text-only generation.

Step 3. Hit run
Lyria 3.5 generates the arrangement, audio, and lyrics in one pass. The track plays in the built-in audio preview. Lyrics appear in the text box beside it.
Ready for: video editors · game engines · DAWs · podcast tools

First time? The default prompt produces a cinematic orchestral track. Hit run as-is to hear what the model can do, then write your own prompt.

RECOMMENDED SETTINGS

Quick-start guide. Find the goal that matches yours and copy the settings.

• Instrumental soundtrack (most people) — Write a prompt with genre, BPM, key, and instruments. Add "instrumental only, no vocals" to skip singing.

• Full song with vocals — Describe the genre and mood. Include verse-chorus-bridge structure in the prompt. The model writes and performs the lyrics.

• Custom lyrics — Paste your own lyrics under Verse and Chorus tags in the prompt. The model sings what you wrote.

• Timed sections — Use timestamps like [0:00–0:30] Intro, [0:30–1:00] Build to control what happens when.

• Music for a photo or scene — Upload an image alongside the prompt. The model reads the mood and lighting and composes a matching track.

• The track is not matching the prompt — Be specific. Name instruments, BPM, and key. "90 BPM, D minor, solo cello" gives more consistent output than "sad music."

• Need a longer track — Write a target length in the prompt, such as "a 2-minute track." Maximum length is about two to three minutes per run.

Prompt: Be specific about genre, instruments, tempo, and structure. "Cinematic orchestral, 90 BPM, D minor, instrumental only" lands closer than "epic movie music." Use timestamps to control arrangement. Over-describing mood without naming instruments makes the output less predictable.

LEARN

📹 Videos

• Intro to Floyo

• ComfyUI 101 Free Course ft. Sebastian Kamph

• Floyo 101 for Team Collaboration

✨ Quick links

• Help docs

• LoRA Training guide

• Discover more workflows

USE CASES

🎬 Filmmakers & Video Editors
Generate a custom soundtrack for a scene instead of digging through stock libraries. Describe the mood, tempo, and instruments and get a track that fits the edit.

🎮 Game Developers
Create background music and ambient audio for levels, menus, and trailers. Run the workflow once per scene to build a full game soundtrack from prompts.

🎙️ Podcasters & Content Creators
Generate intro music, outros, and background beds in the exact style and length you need, without licensing headaches.

🎵 Songwriters & Musicians
Prototype song ideas fast. Describe the genre, structure, and vibe, hear a full arrangement back, then take the idea into a DAW to develop further.

🎨 Concept Artists & Mood Boards
Upload a scene image and generate a matching soundtrack. Useful for pitching a visual concept with audio that sells the atmosphere.

WHAT WORKS BEST / WHAT TO AVOID

✅ Works great

• Specific genre, BPM, and key in the prompt

• Named instruments and arrangement structure

• Timestamped sections for controlled builds

• Instrumental-only requests for film and game scores

⚠️ May produce softer results

• Vague prompts like "make something cool"

• Requests longer than about three minutes

• Expecting to edit or extend a finished track (single-turn only)

• Mixing too many conflicting genres in one prompt

FAQ

What is Lyria 3.5 and who made it?
Lyria 3.5 is Google DeepMind's music generation model. It writes and performs full songs from a text prompt, outputting 44.1 kHz stereo audio alongside lyrics. Google launched it in Flow Music on 29 July 2026, then released it through the Gemini API and Gemini app on 4 September 2026.

How does Lyria 3.5 generate music from text?
You write a prompt describing the genre, instruments, tempo, key, and structure. The model reasons out an arrangement, then generates the audio and lyrics in a single pass. You can use timestamps in the prompt to control which sections play when. The output is a complete stereo track, not a loop or a stem.

Can Lyria 3.5 generate vocals and lyrics?
Yes. If your prompt describes a song, the model writes lyrics and sings them with expressive phrasing. You can also paste your own lyrics in the prompt and the model performs them. To skip vocals entirely, add "instrumental only, no vocals" to the prompt.

How is Lyria 3.5 different from Suno and Udio?
The main difference is how the model was trained. Google states Lyria was trained on licensed content, while Suno and Udio face copyright litigation. Lyria 3.5 also accepts image input to compose music matching a visual scene. The tradeoff: Lyria 3.5 maxes out at about three minutes and is single-turn, so you cannot extend or edit a finished track. Suno supports longer outputs and stem exports.

Does Lyria 3.5 output have a watermark?
Yes. Every track carries a SynthID watermark, Google's system for marking AI-generated audio. The watermark is inaudible and does not affect how the track sounds, but it stays embedded in the file so the audio can be identified as AI-generated.

Can I use Lyria 3.5 music commercially?
Lyria 3.5 is a proprietary Google model, so commercial use is governed by Google's terms rather than an open license. Commercial use is generally allowed under those terms. Check Google's current usage policy, note that outputs carry the SynthID watermark, and confirm you have the rights to any image you upload as input.

How to run Lyria 3.5 online?
You can run Lyria 3.5 online through Floyo. No installation, no setup, no API key to wire up. Open the workflow in your browser, write a prompt, and hit run. Free to try.

WHY FLOYO?

Floyo is the only platform with team collaboration for ComfyUI in the browser. You run workflows with no install. You share run history, assets, and models across your team. You pay only when you generate. Floyo supports open-source and closed-source models.

A designer runs an edit and likes the result. A teammate opens that exact run from shared history and keeps going. No file handoffs. No version confusion.

For studios and enterprise teams, Floyo adds private workspaces, pooled resources, and a team usage dashboard. Other ComfyUI cloud tools run for one person at a time. Floyo runs for the whole team, with transparent per-generation costs.

Ready to try it?
Write a prompt describing the music you want and hit run. The default prompt is already set.

→ Launch Workflow, Free

Questions? Watch the free course or check the FAQ above.